Evaluation of Parallel Paradigms on Anisotropic Nonlinear Diffusion
Anisotropic Nonlinear Diffusion (AND) is a powerful noise reduction technique in the field of computer vision. This method is based on a Partial Differential Equation (PDE) tightly coupled with a massive set of eigensystems. Denoising large 3D images in biomedicine and structural cellular biology by AND is extremely expensive from a computational point of view, and the requirements may become so huge that parallel computing turns out to be essential. This work addresses the parallel implementation of AND. The parallelization is carried out by means of three paradigms: (1) Shared address space paradigm, (2) Message passing paradigm, and (3) Hybrid paradigm. The three parallel approaches have been evaluated on two parallel platforms: (1) a DSM (Distributed Shared Memory) platform based on cc-NUMA memory access and (2) a cluster of Symmetric biprocessors. An analysis of the performance of the three strategies has been accomplished to determine which is the most suitable paradigm for each platform.
KeywordsMessage Passing Structure Tensor Distribute Shared Memory Hybrid Code Parallel Paradigm
Unable to display preview. Download preview PDF.
- 1.Weickert, J.: Anisotropic Diffusion in Image Processing. Teubner (1998)Google Scholar
- 19.Press, W.H., Flannery, B.P., Teukolsky, S.A., Vetterling, W.T.: Numerical Recipes: The Art of Scientific Computing. Cambridge University Press, Cambridge (1992)Google Scholar
- 21.Dunigan, T.H., Vetter, J.S., Worley, P.: Performance evaluation of the SGI Altix 3700. In: Proceedings of the IEEE Intl. Conf. Parallel Processing, ICPP, pp. 231–240 (2005)Google Scholar